Load Balancing Virtual Servers, Services, and Monitors
Load balancing is the highest-yield domain on this exam. It tests virtual server types, service and service-group binding, monitor behavior, load-balancing methods, and persistence. Most questions ask you to explain a state transition or why traffic goes to the wrong place.
Building blocks
- An LB virtual server is a VIP, port, and protocol. Services point to a server IP and port; service groups bundle members that share monitors.
- Monitors define health checks: type, interval, response timeout, retries, and destination. If no monitor is bound, the service is assumed up.
- Load-balancing methods include least connection (the default), round robin, least response time, least bandwidth, source IP hash, and custom methods. Persistence overrides the method for an existing session.
- Down state flush, when enabled, resets existing connections to a service that goes down. A backup virtual server can serve traffic when all primary services are down.
Worked example
A VIP 10.10.0.50:80 fronts three web servers. An HTTP-ECV monitor sends GET /health, expects HTTP 200, runs every 5 seconds, has a 2-second response timeout, and uses 3 retries. Server B begins returning HTTP 500. After three consecutive failed checks, roughly 15 seconds plus timeout time, B is marked down and removed from rotation. If all three fail, the virtual server goes down, and a backup virtual server can return a maintenance page. If down state flush is enabled, existing sessions to B are reset. The reasoning is that the monitor is a control-plane health decision; retries and intervals determine detection time, not the load-balancing method.
Traps that cost marks
Binding a monitor only to the virtual server instead of the service is a common error. A service-group monitor applies to every member. Persistence can keep a user on a failed server until the persistence timeout expires. Least connection can still send traffic to a slow server that has few connections. Round robin ignores server load entirely. Enabling down state flush can cause user-visible resets, and assuming a service is healthy because no monitor is bound is unsafe.